390 JOURNAL OF LAW AND POLICY
I. METHODS
In all cases, the dialogues analyzed here have already been
independently transcribed and are available on the web or in
publications cited. The importance of this, given the intention to
study repetition across successive dialogue turns, is that the turn-
taking structure of dialogue has been determined independently,
without evidence of prior reflection on the possibility of this sort
of analysis being undertaken. Temporally, the transcripts are
partially ordered given that contributions of interlocutors are
interleaved; however, temporal overlap analysis is not
conducted. Ideally, one would have available not just textual
transcripts that indicate the sequence of turns but also the timing
of those turns so that temporal overlap of turns can be taken into
account. However, as with the dialogues analyzed here, one
cannot be guaranteed the availability of timing information.
A decision has to be made with respect to the level of
linguistic description at which to consider repetition
(tokenization): morphemes, words, part-of-speech (“POS”)
labels, concepts, etc. or combinations thereof. The units of
representation decided on are types, and their instances are
tokens. At this stage, punctuation marks are disregarded.
Representation of semantic information that is not directly
lexically encoded is not made, since it is not safe to conclude
that speakers accept as true all logically valid consequences of
their assertions. The text is individuated as words and restricted
part-of-speech labeling. POS labeling is only used for personal
pronouns to capture the fact that, ordinarily, they are not
repeated verbatim but with complementarity, in dialogue that
proceeds successfully. Thus, the sole other treatment of the data
analyzed here (apart from ignoring punctuation) is to transform
dialogues in the form of examples like those numbered below
(1) and (3) into those like (2) and (4), respectively; that is,
complementary first-person and second-person personal
pronouns are replaced with a single item (“IY,” regardless of
grammatical number). No deeper parsing is deployed and no
other POS labels are used; even third-person pronouns are left
intact. Avoiding parsing is desirable to ensure that the methods